Medical image registration method, medical image processing apparatus, storage medium, and program product

Abstract

A medical image registration method, a medical image processing device, a storage medium, and a program product enable registration between two medical images even if a registration target organ or a registration target site included in one of the medical images is largely deformed. The medical image registration method of the present invention, which registers a plurality of medical images including a registration target organ, includes: a target region extraction step of obtaining a registration target region by obtaining an edge of the registration target organ for one of the medical images; a deformation degree calculation step of calculating a deformation degree at a plurality of positions within the registration target region; and an image rigid registration step of rigidly registering the plurality of medical images based on the calculated deformation degree at each position and a similarity between the plurality of medical images at the position.

Description

Technical Field

[0001] The embodiments of the present invention relate to a medical image registration method, a medical image processing device, a storage medium, and a program product. Background Technology

[0002] Common medical images include ultrasound images (US images), computed tomography (CT) images, and magnetic resonance imaging (MR) images. Depending on the needs of different medical procedures, medical image processing devices may require registration between different types of medical images under varying circumstances.

[0003] For example, some surgeries or examinations require the guidance of ultrasound images. However, ultrasound images suffer from low contrast and high noise. Therefore, it is sometimes necessary to fuse ultrasound images with magnetic resonance imaging (MRI) images to more accurately locate certain organs or body parts. In image fusion, a correspondence between the ultrasound coordinate system and the MRI coordinate system is usually established through rigid registration of ultrasound and MRI images.

[0004] However, acquiring ultrasound images requires the probe to be in close contact with the area being scanned. This process can easily cause deformation of the organ being examined, and images containing deformation can negatively impact the accuracy of rigid registration. Especially when the ultrasound image contains significant deformation, if a portion of the contour (edge) of a deformed organ resembles a portion of the contour (edge) of another organ in the MRI image, it may lead to incorrect registration between the ultrasound image and other organs in the MRI image. This results in failure of rigid registration and failure of image fusion. Summary of the Invention

[0005] In view of the problems existing in the prior art, the present invention provides a medical image registration method, a medical image processing device, a storage medium and a program product, which can achieve registration between two medical images even if the organ or part of the object to be registered contained in a certain medical image has large deformation.

[0006] The medical image registration method of the present invention registers multiple medical images containing a registration target organ, comprising: an object region extraction step, wherein for one of the medical images, the edge of the registration target organ is obtained to obtain a registration target region; a deformation degree calculation step, wherein the deformation degree at multiple locations within the registration target region is calculated; and an image rigid registration step, wherein the multiple medical images are rigidly registered based on the calculated deformation degree at each location and the similarity between the multiple medical images at that location.

[0007] Furthermore, the degree of deformation includes at least one of a parameter representing the concavity or convexity of the edge, a parameter representing the smoothness of the edge, and a parameter representing the distance between the organ and the extruder. When the degree of deformation includes two or more parameters, the weights of the two or more coefficients are adjusted according to the deformation of the registered organ to generate an optimal combination of deformation coefficients.

[0008] Furthermore, in the image rigid registration step, a weighted similarity is calculated based on the degree of deformation at each location and the similarity between the various medical images at that location, with more weight assigned to areas with smaller deformation and less weight assigned to areas with larger deformation. Rigid registration is then performed on the various medical images based on this weighted similarity.

[0009] In addition, after the deformation degree calculation step, the following step may be performed: a deformation degree correction step, which corrects the deformation degree according to the magnitude of the vector in the deformation vector field corresponding to the two medical images, such that the larger the magnitude, the greater the deformation degree.

[0010] The medical image processing apparatus of the present invention is used to register multiple medical images containing a registration target organ, comprising: a target region extraction unit, for one of the medical images, obtaining the edge of the registration target organ to obtain a registration target region; a deformation degree calculation unit, calculating the deformation degree at multiple locations within the registration target region; and an image rigid registration unit, performing rigid registration on the multiple medical images based on the calculated deformation degree at each location and the similarity between the multiple medical images at that location.

[0011] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the medical image registration method.

[0012] The computer program product of the present invention includes a computer program, characterized in that, when the computer program is executed by a processor, it implements the steps of the medical image registration method.

[0013] Invention Effects

[0014] According to the medical image registration method, medical image processing device, storage medium and program product described above, by calculating the degree of deformation at multiple locations in the medical image, and performing rigid registration based on the calculated degree of deformation at each location during the registration process, even if the organ or part of the registration object contained in a certain medical image has a large deformation, the interference of the large deformation in the medical image on the rigid registration can be suppressed, thereby improving the accuracy of rigid registration between two medical images.

[0015] Furthermore, by calculating the weighted similarity and performing rigid registration by assigning more weight to regions with smaller deformations and less weight to regions with larger deformations, the adverse effects of regions with large deformations on the registration results can be automatically reduced, and the contribution of regions with no deformation or small deformations to the registration results can be strengthened. This more reliably suppresses the interference of large deformations in medical images on rigid registration.

[0016] Furthermore, by selecting at least one of the edge concavity / convexity parameter, edge smoothness parameter, and probe distance parameter, deformation parameters adapted to the characteristics of the target organ can be used, improving the accuracy of the calculated deformation degree. Moreover, by adjusting the weights of the two or more coefficients according to the deformation of the target organ to generate an optimal combination of deformation coefficients, deformation parameters more adapted to the characteristics of the target organ can be obtained.

[0017] Furthermore, the deformation vector field was obtained through non-rigid registration, and the deformation parameters or the optimal combination of deformation coefficients were corrected by using the magnitude of the deformation represented by the modulus of each vector, thereby obtaining more accurate deformation parameters or the optimal combination of deformation coefficients. Attached Figure Description

[0018] Figure 1 This is a functional block diagram illustrating the medical image processing apparatus of the first embodiment.

[0019] Figure 2 This is a flowchart illustrating the registration process of a medical image according to the first embodiment.

[0020] Figure 3 This is a schematic diagram illustrating the extraction of the prostate region (registration target region) from an ultrasound image.

[0021] Figure 4 This is a schematic diagram showing the shape of the prostate region in a magnetic resonance imaging (MRI) image.

[0022] Figure 5 It is a diagram used to illustrate the convexity and concavity of the prostate margins in an ultrasound image.

[0023] Figure 6 This is a schematic diagram used to illustrate the relationship between the edge curvature and deformation of the prostate.

[0024] Figure 7 This is a schematic diagram used to illustrate the relationship between probe distance and deformation in ultrasound images.

[0025] Figure 8 This is a schematic diagram illustrating a successful registration example of an ultrasound image and a magnetic resonance image according to the first embodiment.

[0026] Figure 9This is a schematic diagram illustrating a comparative example of a registration failure between an ultrasound image and a magnetic resonance image.

[0027] Figure 10 This is a diagram showing an example of registration reliability.

[0028] Figure 11 This is a functional block diagram illustrating the medical image processing apparatus of the second embodiment.

[0029] Figure 12 This is a flowchart illustrating the registration of medical images according to the second embodiment.

[0030] Figure 13 This is a schematic diagram illustrating the relationship between vectors and the degree of deformation in the deformation vector field of the second embodiment. Detailed Implementation

[0031] (First Implementation)

[0032] The medical image processing apparatus 10 according to the first embodiment of the present invention will now be described.

[0033] Figure 1 This is a functional block diagram illustrating the medical image processing apparatus 10 according to the first embodiment. For example... Figure 1 As shown, the medical image processing device 10 includes a storage unit 100, an image acquisition unit 200, a feature extraction unit 300, an object region extraction unit 400, a deformation degree calculation unit 500, an image rigid registration unit 600, and a display unit 700.

[0034] The storage unit 100 in this invention stores various types of data. Specifically, the storage unit 100 stores at least various medical images used for image registration, as well as fused images obtained after registration. The storage unit 100 is implemented, for example, using semiconductor memory elements such as RAM (Random Access Memory), flash memory, hard disks, optical disks, etc.

[0035] The image acquisition unit 200, feature extraction unit 300, object region extraction unit 400, deformation degree calculation unit 500, and image rigid registration unit 600 in this invention are implemented, for example, by a processor. The term "processor" can refer to a CPU (Central Processing Unit), GPU (Graphics Processing Unit), or an application-specific integrated circuit (ASIC), programmable logic device (e.g., a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)).

[0036] Furthermore, each of the aforementioned units is stored in the storage unit 100 in the format of a computer-executable program. The processor reads each program from the storage unit 100 and executes the read program to implement the function corresponding to each program. Alternatively, the program may not be stored in the storage unit 100, but may be loaded directly into the processor's circuitry. In this case, the processor implements the function by reading and executing the program loaded into the circuitry.

[0037] Figure 1 The medical image processing apparatus 10 shown here only illustrates the functional components relevant to the present invention. Although not illustrated, the medical image processing apparatus 10 may also have other image processing functions. Furthermore, the medical image processing apparatus 10 may also be implemented as part of the functions of an ultrasound imaging device, a CT imaging device, a magnetic resonance imaging device, or a multifunctional imaging device.

[0038] The image acquisition unit 200 acquires two or more medical images as registration target images. The registration target images can be images acquired from the storage unit 100 or images acquired during surgery by imaging equipment such as ultrasound imaging equipment. For example, in the case of prostate biopsy, it is necessary to acquire a magnetic resonance image (3D image) of the prostate in advance and store it in the storage unit 100, and then acquire an ultrasound image (3D image) after the surgery begins.

[0039] The feature extraction unit 300 extracts image features from the image to be registered. The feature maps obtained by extracting image features include, for example, grayscale images, grayscale histograms, variance maps, gradient maps, etc.

[0040] The object region extraction unit 400, for one type of medical image (e.g., ultrasound image) among various medical images, acquires the edges of the registration target organ and extracts the registration target region corresponding to the registration target organ from the medical image. This extraction can be performed using existing edge segmentation algorithms or deep learning region extraction algorithms.

[0041] The deformation degree calculation unit 500 calculates the deformation degree at each location within the registration target area. In this embodiment, a "location" can be either a single pixel within the registration target area or a sub-region composed of multiple pixels. That is, the deformation degree can be calculated for each pixel or for each sub-region. The deformation degree includes at least one of the following: a parameter representing the concavity / convexity of the edge (i.e., edge concavity / convexity parameter), a parameter representing the smoothness of the edge (i.e., edge smoothness parameter), and a parameter representing the distance between the organ and the extruded material (i.e., probe distance parameter). The calculation method for the deformation degree will be explained later.

[0042] The image rigid registration unit 600 performs rigid registration on various medical images based on the calculated degree of deformation at each location. During rigid registration, the optimal similarity position is found by adjusting the rotation and translation amounts between the images. To reduce the impact of deformation of the target organ on registration, a deformation coefficient is added to the similarity calculation, assigning greater weight to regions with smaller deformations and less weight to regions with larger deformations.

[0043] The display unit 700 displays the fused image as the registration result. Furthermore, it is preferable to display the registration reliability of different positions in the registration result based on the similarity of multiple positions in the fused image while displaying the fused image. As a display method, different colors can be used to represent the level of reliability for different positions in the fused image. By marking positions with small deformation and high registration reliability and positions with large deformation and low registration reliability in the fused image with different colors, the reliability of each position in the fused image can be intuitively indicated to the user.

[0044] The following explanation uses the registration of ultrasound and magnetic resonance images containing the prostate as an example.

[0045] Figure 2 This is a flowchart illustrating the registration process of a medical image according to the first embodiment.

[0046] First, in the image acquisition step S11, the image acquisition unit 200 acquires an ultrasound image containing the prostate and a magnetic resonance image containing the prostate as registration target images. In this embodiment, the ultrasound image is an image acquired by an ultrasound imaging device after the start of surgery, and the magnetic resonance image is an image acquired in advance by a magnetic resonance imaging device and stored in the storage unit 100.

[0047] Next, in the feature extraction step S12, the feature extraction unit 300 extracts image features for the two registration object images. In this embodiment, the extracted image features are gradient maps, thereby obtaining the gradient images of each of the two registration object images.

[0048] Next, in the object region extraction step S13, the object region extraction unit 400 obtains the edge of the prostate as the registration target organ from the ultrasound image, and extracts the prostate region (registration target region) corresponding to the prostate from the ultrasound image. The extraction method used in this embodiment is, for example, an edge segmentation algorithm.

[0049] Figure 3 This is a schematic diagram illustrating the extraction of the prostate region (registration target region) from an ultrasound image. Figure 3 The following is an example using an axial view. The upper part is a schematic diagram of the ultrasound image without the prostate region being extracted, and the lower part is a schematic diagram of the ultrasound image after the prostate region has been extracted.

[0050] Next, in the deformation degree calculation step S14, the deformation degree calculation unit 500 calculates the deformation degree of each location in the prostate region of the ultrasound image. In this embodiment, three parameters are used: edge concavity / convexity parameter, edge smoothness parameter, and probe distance parameter, to calculate the deformation degree of each location. Furthermore, the weights of these three deformation coefficients are adjusted according to the deformation of the prostate in the ultrasound image, thereby generating an optimal combination of deformation coefficients.

[0051] The edge concavity / convexity parameters are explained below.

[0052] The reason for using edge roughness to calculate the degree of deformation is that, in the case where the prostate, which is the registration target organ in this embodiment, is not compressed, such as Figure 4 As shown in the axial view of the magnetic resonance image, the overall outline of the organ in its normal state is roughly chestnut-shaped, but when the prostate is squeezed by an ultrasound probe, its edge irregularities change.

[0053] When calculating the edge concavity / convexity coefficient, the positive direction of the organ contour is first determined. In this embodiment, the positive direction of the prostate is the direction in which the contour bulges outward. At this time, the edge concavity / convexity coefficient d1(p) can be calculated by the following equation 1.

[0054]

[0055] Where p is any point in the prostate region, q is the edge point closest to p, convex(q) is the concavity of q (1 for the positive direction of outward protrusion and 0 for the negative direction of inward concavity), distance(p,q) is the distance between p and q, T is a preset distance threshold, and C is a preset fixed coefficient.

[0056] As can be seen from Equation 1 above, the closer point p is to the edge, the greater the influence of the edge's concavity and convexity; when the distance between point p and point q is greater than the distance threshold T, the edge's concavity and convexity no longer affect the deformation coefficient.

[0057] Figure 5 It is a diagram used to illustrate the convexity and concavity of the prostate margin in an ultrasound image. Figure 5 The area within the dashed box represents the portion of the prostate gland's edge that has undergone significant deformation due to compression by the ultrasound probe. Compared to its normal shape, the irregularity of this portion has changed. Specifically, from the outward-protruding direction (refer to...) Figure 4 The shape changed to a negative direction, concave inwards. On the other hand, the portion of the prostate's edge other than the dashed box is the least deformed part, and its convexity and concavity do not change significantly compared to its normal shape.

[0058] Next, the edge smoothness parameters will be explained.

[0059] The reason for calculating the degree of deformation by edge smoothness is that, when the prostate, the organ to be registered, is not compressed, the edge of the organ is smooth and has a small curvature in its normal state. When it is compressed by the ultrasound probe and deformed, the curvature of its edge will increase.

[0060] When calculating the edge smoothness coefficient, the edge curvature is included in the deformation coefficient. The edge smoothness coefficient d2(p) can then be calculated using Equation 2 below.

[0061]

[0062] Where p is any point in the prostate region, q is the nearest edge point to p, curvature(q) is the curvature of q, distance(p,q) is the distance between p and q, T is a pre-set distance threshold, and C is a pre-set fixed coefficient.

[0063] As can be seen from Equation 2 above, the closer point p is to the edge, the greater the influence of the edge curvature. When the distance between point p and point q is greater than the distance threshold T, the edge curvature no longer affects the deformation coefficient.

[0064] Figure 6 This is a schematic diagram used to illustrate the relationship between the edge curvature and deformation of the prostate. Figure 6 The area within the dotted box in the lower middle section represents the portion of the prostate gland that has undergone significant deformation due to compression by the ultrasound probe. Compared to its normal shape, the smoothness of this portion has changed, meaning its curvature has increased. On the other hand, Figure 6 The area within the dotted box at the top center represents the portion of the prostate gland's edge that experiences minimal deformation after being compressed by the ultrasound probe. Compared to its normal shape, this portion shows little change in smoothness and curvature. Furthermore, in Figure 6 As can be seen, the edge curvature within the lower dashed box is greater than that within the upper dashed box.

[0065] Next, the probe distance parameters will be explained.

[0066] The reason for calculating the degree of deformation by measuring the distance from the probe is that the area closer to the probe is more likely to be compressed and deformed.

[0067] When calculating the probe distance coefficient, a larger deformation coefficient is assigned to areas closer to the probe, and a smaller deformation coefficient is assigned to areas farther from the probe. In this case, the probe distance coefficient d3(p) can be calculated using Equation 3 below.

[0068] d(p)=decay_function(distance(p,probe)) (Formula 3)

[0069] Where p is any point in the prostate region, distance(p, probe) is the distance between point p and the probe, and decay_function is the decay function. The decay function decay_function can be an exponential decay function. or power decay function x k , where k is the attenuation parameter.

[0070] As can be seen from Equation 3 above, the closer point p is to the probe, the greater its influence on the deformation coefficient; the farther point p is from the probe, the smaller its influence on the deformation coefficient.

[0071] Figure 7 This is a schematic diagram used to illustrate the relationship between probe distance and deformation in ultrasound images. Figure 7The first sector with a smaller radius is the part closer to the ultrasonic probe, and this part undergoes greater deformation. Figure 7 The second sector, located outside the first sector, is the part farther from the ultrasonic probe, and its deformation is smaller.

[0072] In this embodiment, the optimal combination of deformation coefficients is generated by adjusting the weights of the three deformation coefficients mentioned above. The optimal combination of deformation coefficients d(p) can be calculated using the following equation 4.

[0073]

[0074] Where, d i (p) is the i-th deformation coefficient, w i This represents the weight corresponding to the i-th deformation coefficient. The proportion of each deformation coefficient can be preset based on empirical values ​​or determined based on machine learning results. For example, the weights of the edge concavity / convexity coefficient, edge smoothness coefficient, and probe distance coefficient are 0.3, 0.3, and 0.4, respectively.

[0075] Next, return to Figure 2 Continuing the explanation. In the image rigid registration step S15, the ultrasonic image and magnetic resonance image in this embodiment are rigidly registered according to the degree of deformation at each location.

[0076] During the registration process, the optimal similarity position is found by adjusting the rotation and translation between images. To reduce the impact of deformation of the organs in the registration object on the registration, the deformation coefficient is added to the similarity calculation. Regions with smaller deformation are assigned a larger weight, and regions with larger deformation are assigned a smaller weight, thus obtaining a weighted similarity.

[0077] The weighted similarity considering the deformation coefficient can be calculated using Equation 5 below.

[0078] similarity_w(p,q)=similarity(p,q)·d(p) -1 (Equation 5)

[0079] Where p is any point on the ultrasound image, q is any point on the magnetic resonance image, similarity(p,q) is the similarity between point p and point q, and d(p) is the deformation coefficient of point p calculated by step S14 of the deformation degree calculation.

[0080] More specifically, similarity(p,q) represents the degree of similarity between any point p in the feature map of the ultrasound image and any point q in the feature map of the magnetic resonance image. Since gradient images of the ultrasound image and the magnetic resonance image are obtained in the feature extraction step S12 in this embodiment, the similarity(p,q) here refers, for example, to the angle between the gradient feature vectors of point p in the gradient image of the ultrasound image and point q in the gradient image of the magnetic resonance image.

[0081] Let the gradient eigenvector at point p be... The gradient eigenvector at point q is In the case of [condition], the similarity between two gradient images can be calculated using Equation 6 below.

[0082]

[0083] Furthermore, when the feature map is a grayscale image instead of a gradient image, the similarity (p,q) refers, for example, to the mutual information between point p in the grayscale image of an ultrasound image and point q in the grayscale image of a magnetic resonance image.

[0084] Furthermore, when the feature map is a variance map instead of a gradient image, the similarity (p,q) refers to, for example, the correlation coefficient between point p in the variance map of an ultrasound image and point q in the variance map of a magnetic resonance image.

[0085] Figure 8 This is a schematic diagram illustrating a successful registration example of an ultrasound image and a magnetic resonance image according to the first embodiment. Figure 8 The following is an example of an illustration using a sagittal view including the prostate.

[0086] exist Figure 8 In the ultrasound image, prostate region A is the prostate region deformed due to the compression of the ultrasound probe. More specifically, the lower right part of the prostate region is more deformed due to the compression of the ultrasound probe. The more deformed part (the part within the dashed box) changes from a shape that protrudes outward before being compressed to a shape that is concave inward.

[0087] In contrast, region B of the prostate in the MRI image is a prostate region that has not undergone deformation. Additionally, region C of the rectum is also shown in the MRI image.

[0088] When registering ultrasound images and magnetic resonance images through the steps of this embodiment, since the deformation degree of each location in the ultrasound image is calculated in the deformation degree calculation step S14, and regions with smaller deformation degrees are given greater weight and regions with larger deformation degrees are given less weight when calculating similarity in the image rigid registration step S15, successful registration is achieved even if there are regions with large deformation in the registration target area, thereby obtaining... Figure 8 The fused image in the image represents the registration result. Figure 8 In the ultrasound image, prostate region A (represented by a single-dotted line) and prostate region B (represented by a solid line) in the magnetic resonance image largely overlap, especially the part with less deformation (the upper half of the roughly elliptical prostate region in the fused image) which has a high degree of overlap.

[0089] As a comparative example Figure 9 The diagram illustrates an example of registration failure between an ultrasound image and a magnetic resonance imaging (MRI) image. In this case, because a significantly deformed portion in the ultrasound image resembles a part of the edge of the rectal region C in the MRI image (the dashed box portion in the MRI image), and since the comparative example performed registration without considering the degree of deformation at each location in the images, a registration failure may occur. Figure 9 The registration result is shown in the fused image.

[0090] exist Figure 9 In the fused images, because the edges of the more deformed portion of prostate region A in the ultrasound image are highly similar to a portion of the edge of rectal region C (the dashed box in the MRI image), the two are incorrectly registered together. Conversely, the edges of prostate region A in the ultrasound image (represented by a dashed line) and prostate region B in the MRI image (represented by a solid line) hardly overlap; instead, they are offset by a certain distance. Figure 9 The fused image in the image is a fused image that failed to be registered.

[0091] Next, return to Figure 2 Continuing the explanation. In display step S16, the display unit is made to display... Figure 8 The fused image in the image is the result of registration.

[0092] Furthermore, in the display step S16, it is preferable to display the registration reliability of different positions in the registration result based on the similarity of multiple positions in the fused image while displaying the fused image. For example, different colors can be used to represent the level of reliability for different positions in the fused image.

[0093] Figure 10This is a diagram illustrating an example of registration reliability. In this fused image, areas with small deformation and high registration reliability are marked in green (represented by a diamond pattern in the diagram), while areas with large deformation and low registration reliability are marked in red (represented by a diagonal line pattern in the diagram). This visually indicates to the user the reliability of different areas in the fused image. In addition to using color, different graphics or patterns can also be used for differentiation.

[0094] The technical effects of the medical image processing apparatus 10 and the medical image registration method according to the first embodiment will be explained below.

[0095] According to the medical image processing apparatus 10 and the medical image registration method of the first embodiment, by calculating the degree of deformation at multiple locations in the ultrasound image and performing rigid registration based on the calculated degree of deformation at each location during the registration process, even if the deformation of the registration target organ or registration target part contained in the ultrasound image is large, the interference of the large deformation on the rigid registration can be suppressed, thereby improving the accuracy of rigid registration between the ultrasound image and the magnetic resonance image.

[0096] Furthermore, by calculating the weighted similarity and performing rigid registration by assigning more weight to regions with smaller deformations and less weight to regions with larger deformations, the adverse effects of regions with large deformations on the registration results can be automatically reduced, and the contribution of regions with no deformation or small deformations to the registration results can be strengthened. This more reliably suppresses the interference of large deformations in ultrasonic images on rigid registration.

[0097] Furthermore, by selecting at least one of the edge concavity / convexity parameter, edge smoothness parameter, and probe distance parameter, deformation parameters adapted to the characteristics of the target organ can be used, improving the accuracy of the calculated deformation degree. Moreover, by adjusting the weights of the two or more coefficients according to the deformation of the target organ to generate an optimal combination of deformation coefficients, deformation parameters more adapted to the characteristics of the target organ can be obtained.

[0098] (Second Implementation)

[0099] Below, refer to Figures 11-13 The medical image processing apparatus 20 of the second embodiment of the present invention will be described.

[0100] Figure 11 This is a functional block diagram illustrating the medical image processing apparatus 20 according to the second embodiment. For example... Figure 11 As shown, the medical image processing device 20 includes, in addition to Figure 1In addition to the units shown, the system also includes a deformation vector field generation unit 800 and a deformation degree correction unit 900. The descriptions identical to those in the first embodiment will be omitted hereafter; only the differences will be explained.

[0101] The deformation vector field generation unit 800 performs non-rigid registration on two medical images to generate the deformation vector field. The deformation vector field between the two medical images can be obtained using existing non-rigid registration algorithms (such as the ICP algorithm).

[0102] The deformation degree correction unit 900 corrects the deformation degree calculated by the deformation degree calculation unit 500 based on the magnitude of the vector in the deformation vector field corresponding to the two medical images, in a manner that the greater the magnitude, the greater the deformation degree.

[0103] Furthermore, the image rigid registration unit 600 in the second embodiment also performs rigid registration on multiple medical images again based on the degree of deformation at each corrected position and the similarity between the two medical images at that position.

[0104] Figure 12 This is a flowchart illustrating the registration of medical images according to the second embodiment.

[0105] exist Figure 12 In the process, after the image rigid registration step S15, a deformation vector field generation step S17, a deformation degree correction step S18, and an image re-rigid registration step S19 are added.

[0106] In step S17, the deformation vector field is generated by non-rigid registration of the ultrasound image and the magnetic resonance image. Existing non-rigid registration algorithms (such as the ICP algorithm) can be used to obtain the deformation vector field from the ultrasound image to the magnetic resonance image.

[0107] Figure 13 This is a schematic diagram illustrating the relationship between vectors and the degree of deformation in the deformation vector field of the second embodiment. For example... Figure 13 As shown, in regions with small deformation (i.e., the upper part of the prostate region), the magnitudes of the vectors are smaller, while in regions with large deformation (i.e., the lower part of the prostate region), the magnitudes of the vectors (especially the bottom few vectors) are larger.

[0108] Next, in the deformation degree correction step S18, the previously calculated deformation degree is corrected according to the magnitude of the vector in the deformation vector field corresponding to the ultrasonic image and the magnetic resonance image, in such a way that the larger the magnitude, the greater the deformation degree.

[0109] As an example of a correction method, the optimal combination of deformation coefficients d(p) calculated in the deformation degree calculation step S14 is corrected by the following equation 7.

[0110] d(p)=d(p)+λ·|transform_field(p)| (Formula 7)

[0111] Where p is any point in the ultrasound image, |transform_field(p)| is the magnitude of the deformation vector field at point p, and λ is the weighting coefficient. The weighting coefficient λ can be preset based on empirical values ​​or based on the results of machine learning, etc. For example, the weighting coefficient λ can be selected in the range of [0.01 to 0.2].

[0112] Next, in the image re-rigid registration step S19, the ultrasound image and the magnetic resonance image are rigidly registered again based on the degree of deformation at each corrected position and the similarity between the two medical images at that position. The image re-rigid registration step S19, based on the first image rigid registration, recalculates the weighted similarity based on the corrected degree of deformation, and further finds the position of optimal similarity by adjusting the rotation and translation amounts between the two medical images within a small range.

[0113] According to the medical image processing apparatus 20 of the second embodiment, a deformation vector field is obtained through non-rigid registration, and the optimal deformation coefficient combination d(p) is corrected by using the magnitude of the deformation represented by the magnitude of each vector, thereby obtaining a more accurate optimal deformation coefficient combination d(p) and further improving the accuracy of image registration.

[0114] (Other variations)

[0115] Several embodiments of the present invention have been described, but these embodiments are merely illustrative and are not intended to limit the scope of the invention. These embodiments can be implemented in various other ways, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, and are also included in the scope of the invention and its equivalents described in the technical solution.

[0116] For example, the above embodiments are illustrated using the registration of ultrasound images and magnetic resonance images containing the prostate as an example, but it can also be applied to the registration of ultrasound images and CT images.

[0117] Furthermore, in the first and second embodiments described above, the fused image is displayed in the display step S16 as the registration result of the image rigid registration step S15 or the image re-rigid registration step S19. In reality, the display step is not limited to displaying the fused image; the fused image may also be further fused with other medical images before being displayed. Alternatively, the fused image may not be displayed; instead, further analysis and processing of the fused image may be performed, and the results of the analysis and processing may be displayed.

[0118] Alternatively, the display step S16 can be omitted. As the registration result of the image rigid registration step S15 or the image re-rigid registration step S19, the correspondence between the ultrasonic image coordinate system and the magnetic resonance image coordinate system can be obtained. This correspondence can be saved and used, for example, in the form of a transformation matrix.

Claims

1. A medical image registration method, comprising registering two medical images containing the organs to be registered, characterized in that, Perform the following steps: The object region extraction step involves obtaining the edges of the organ to be registered, based on one type of medical image from the medical images, to obtain the registration object region. The deformation degree calculation step involves calculating the deformation degree at multiple locations within the registration target area, including at least one of the following parameters: parameters representing the concavity / convexity of the edge, parameters representing the smoothness of the edge, and parameters representing the distance between the organ and the extrusion material. If the deformation degree includes two or more parameters, the weights of the two or more coefficients are adjusted according to the deformation of the registration target organ to generate an optimal combination of deformation coefficients. The image rigid registration step involves calculating a weighted similarity based on the calculated degree of deformation at each location and the similarity between the two medical images at that location. This weights regions with smaller deformations are assigned more weight, while regions with larger deformations are assigned less weight. Rigid registration of the two medical images is then performed based on this weighted similarity. After the image rigid registration step, the following steps are performed: The deformation vector field generation step involves performing non-rigid registration on the two medical images to generate a deformation vector field. The deformation degree correction step involves correcting the deformation degree based on the magnitude of the vectors in the deformation vector field corresponding to the two types of medical images, in a manner that a larger magnitude indicates a greater degree of deformation. The image re-registration step involves rigidly registering the two medical images again based on the degree of deformation at each corrected location and the similarity between the two medical images at that location. In the image rigid registration step, the weighted similarity, which takes into account the deformation coefficients representing the degree of deformation, is calculated using the following formula: Where p is any point on the medical image, q is any point on a medical image other than the medical image, similarity(p, q) is the similarity between point p and point q, and d(p) is the deformation coefficient of point p calculated by the deformation degree calculation step.

2. The medical image registration method according to claim 1, characterized in that, After the image rigid registration step, the following steps are performed: The display step shows the fused image as the registration result. In the display step, the registration reliability of different positions in the registration result is displayed based on the similarity of the multiple positions in the fused image.

3. The medical image registration method according to claim 1, characterized in that, Each of the plurality of locations is a pixel within the registration target area, or a sub-region composed of a plurality of pixels.

4. The medical image registration method according to claim 2, characterized in that, In the display step, different colors are used to represent the level of reliability for different locations in the fused image.

5. The medical image registration method according to claim 1, characterized in that, The medical image mentioned is an ultrasound image.

6. A medical image processing apparatus for registering two medical images containing organs to be registered, characterized in that, have: The object region extraction unit, for one of the medical images, obtains the edge of the organ to be registered to obtain the registration object region; The deformation degree calculation unit calculates the deformation degree of multiple locations within the registration target area, including at least one of the parameters representing the concavity and convexity of the edge, the smoothness of the edge, and the distance of the organ from the extruder. When the deformation degree includes two or more parameters, the unit adjusts the weights of the two or more coefficients according to the deformation of the registration target organ to generate an optimal combination of deformation coefficients. The image rigid registration unit calculates a weighted similarity based on the degree of deformation at each location and the similarity between the two medical images at that location, assigning more weight to areas with smaller deformation and less weight to areas with larger deformation, and then performs rigid registration on the two medical images based on the weighted similarity. The deformation vector field generation unit performs non-rigid registration on the two medical images after the rigid registration unit has performed rigid registration, and generates a deformation vector field. as well as The deformation degree correction unit corrects the deformation degree based on the magnitude of the vectors in the deformation vector field corresponding to the two types of medical images, in a manner that a larger magnitude indicates a greater degree of deformation. The image rigid registration unit performs rigid registration on the two medical images again based on the degree of deformation at each corrected position and the similarity between the two medical images at that position. In the image rigid registration unit, the weighted similarity, which takes into account the deformation coefficient representing the degree of deformation, is calculated by the following formula: Where p is any point on the medical image, q is any point on a medical image other than the medical image, similarity(p, q) is the similarity between point p and point q, and d(p) is the deformation coefficient of point p calculated by the deformation degree calculation unit.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the medical image registration method of claim 1.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the medical image registration method of claim 1.

Images